search

apache / pulsar

Apache Pulsar - distributed pub-sub messaging system
https://pulsar.apache.org/
Apache License 2.0
14.23k stars 3.58k forks source link

[Bug] Pulsar-Worker - failed to load functions #19860

Closed ntilivea closed 1 year ago

ntilivea commented 1 year ago

Search before asking

Version

Pulsar version 2.10.2

Minimal reproduce step

Set-up a cluster with dedicated Pulsar Workers / 5 Zookeeper nodes & Brokers/Bookies/Proxies as needed - each of these are separate Linux machines.

Try to upload a function to the worker using : bin/pulsar-admin --admin-url https://localhost:6751/ functions create --jar /home/user/pulsar-functions.jar --function-config-file /home/user/deduplication-function-config.yml

Received an error :

Feb 23 11:22:41 pulsar-worker-0005 pulsar[8282]: 11:22:41.499 [function-web-25-7] ERROR org.apache.pulsar.functions.worker.rest.api.FunctionsImpl - Failed process Function public/default/TokenGenerationInfoDeduplicationFunction package: Feb 23 11:22:41 pulsar-worker-0005 pulsar[8282]: org.apache.distributedlog.exceptions.WriteException: Write rejected because stream public/default/TokenGenerationInfoDeduplicationFunction/5cf0765e-a677-47d1-a7aa-1f76ec8c5b19-pulsar-functions.jar has encountered an error : writer has been closed due to error. .....

further investigation shown that the issue comes from :

Feb 23 11:22:38 pulsar-worker-0005 pulsar[8282]: java.net.ConnectException: Connection refused Feb 23 11:22:38 pulsar-worker-0005 pulsar[8282]: at sun.nio.ch.Net.pollConnect(Native Method) ~[?:?] Feb 23 11:22:38 pulsar-worker-0005 pulsar[8282]: at sun.nio.ch.Net.pollConnectNow(Net.java:672) ~[?:?] Feb 23 11:22:38 pulsar-worker-0005 pulsar[8282]: at sun.nio.ch.SocketChannelImpl.finishConnect(SocketChannelImpl.java:946) ~[?:?] Feb 23 11:22:38 pulsar-worker-0005 pulsar[8282]: at org.apache.zookeeper.ClientCnxnSocketNIO.doTransport(ClientCnxnSocketNIO.java:344) ~[org.apache.zookeeper-zookeeper-3.6.3.jar:3.6.3] Feb 23 11:22:38 pulsar-worker-0005 pulsar[8282]: at org.apache.zookeeper.ClientCnxn$SendThread.run(ClientCnxn.java:1290) ~[org.apache.zookeeper-zookeeper-3.6.3.jar:3.6.3] Feb 23 11:22:39 pulsar-worker-0005 pulsar[8282]: 11:22:39.548 [function-web-25-7-SendThread(localhost:2181)] INFO org.apache.zookeeper.ClientCnxn - Opening socket connection to server localhost/127.0.0.1:2181.

Seems that the worker tries to load the jar of the functions to the localhost. This is because the entry in the remote Zookeper (/apache-zookeeper-3.7.1-bin/bin/zkCli.sh get /pulsar/functions) looks like :

1 BKDL {"1":{"str":"localhost:2181"},"2":{"str":"/ledgers"},"3":{"tf":1},"4":{"tf":0},"5":{"str":"localhost:2181"},"6":{"str":"localhost:2181"},"7":{"str":"localhost:2181"}}

Although the settings from the functions_worker.yml & is correctly read by the worker :

Feb 23 08:30:42 pulsar-worker-0005 pulsar[8282]: 08:30:42.614 [main] INFO org.apache.pulsar.functions.worker.PulsarWorkerService - Worker Configs: WorkerConfig(workerId=worker-102.10.14.145, workerHostname=102.10.14.145, workerPort=6750, workerPortTls=6751, authenticateMetricsEndpoint=false, includeStandardPrometheusMetrics=false, jvmGCMetricsLoggerClassName=null, numHttpServerThreads=8, httpRequestsLimitEnabled=false, httpRequestsMaxPerSecond=100.0, configurationStoreServers=localhost:2181, configurationMetadataStoreUrl=zk:102.10.14.149:2181,102.10.14.229:2181,102.10.21.152:2181,102.10.14.148:2181,102.10.14.230:2181,

and used :

Feb 23 08:30:42 pulsar-worker-0005 pulsar[8282]: 08:30:42.624 [main] INFO org.apache.distributedlog.api.namespace.NamespaceBuilder - No feature provider is set. All features are disabled now. Feb 23 08:30:42 pulsar-worker-0005 pulsar[8282]: 08:30:42.632 [main] INFO org.apache.distributedlog.impl.BKNamespaceDriver - Created shared zooKeeper client builder dlzk:distributedlog://102.10.14.149:2181,102.10.14.229:2181,102.10.21.152:2181,102.10.14.148:2181,102.10.14.230:2181/pulsar/functions:factory_writer_shared: zkServers = 102.10.14.149:2181,102.10.14.229:2181,102.10.21.152:2181,102.10.14.148:2181,102.10.14.230:2181, numRetries = 3, sessionTimeout = 30000, retryBackoff = 5000, maxRetryBackoff = 30000, zkAclId = null. Feb 23 08:30:42 pulsar-worker-0005 pulsar[8282]: 08:30:42.635 [main] INFO org.apache.zookeeper.ZooKeeper - Initiating client connection, connectString=102.10.14.149:2181,102.10.14.229:2181,102.10.21.152:2181,102.10.14.148:2181,102.10.14.230:2181 sessionTimeout=30000 watcher=org.apache.bookkeeper.zookeeper.ZooKeeperWatcherBase@47fce2c4 Feb 23 08:30:42 pulsar-worker-0005 pulsar[8282]: 08:30:42.636 [main] INFO org.apache.zookeeper.ClientCnxnSocket - jute.maxbuffer value is 10485760 Bytes Feb 23 08:30:42 pulsar-worker-0005 pulsar[8282]: 08:30:42.636 [main] INFO org.apache.zookeeper.ClientCnxn - zookeeper.request.timeout value is 0. feature enabled=false Feb 23 08:30:42 pulsar-worker-0005 pulsar[8282]: 08:30:42.636 [main-SendThread(102.10.14.229:2181)] INFO org.apache.zookeeper.ClientCnxn - Opening socket connection to server ip-102-10-14-229.someplace/102.10.14.229:2181. Feb 23 08:30:42 pulsar-worker-0005 pulsar[8282]: 08:30:42.637 [main-SendThread(102.10.14.229:2181)] INFO org.apache.zookeeper.ClientCnxn - SASL config status: Will not attempt to authenticate using SASL (unknown error) Feb 23 08:30:42 pulsar-worker-0005 pulsar[8282]: 08:30:42.637 [main-SendThread(102.10.14.229:2181)] INFO org.apache.zookeeper.ClientCnxn - Socket connection established, initiating session, client: /102.10.14.145:59718, server: ip-102-10-14-229.someplace/102.10.14.229:2181 Feb 23 08:30:42 pulsar-worker-0005 pulsar[8282]: 08:30:42.640 [main-SendThread(102.10.14.229:2181)] INFO org.apache.zookeeper.ClientCnxn - Session establishment complete on server ip-102-10-14-229.someplace/102.10.14.229:2181, session id = 0x20000049e550004, negotiated timeout = 30000

What did you expect to see?

Functions to be corrected uploaded and to be able to process messages

What did you see instead?

Error creating. Faulty Zookeper entry :

/apache-zookeeper-3.7.1-bin/bin/zkCli.sh get /pulsar/functions

1 BKDL {"1":{"str":"localhost:2181"},"2":{"str":"/ledgers"},"3":{"tf":1},"4":{"tf":0},"5":{"str":"localhost:2181"},"6":{"str":"localhost:2181"},"7":{"str":"localhost:2181"}}

Anything else?

Workaround : manually updating the Zookeper entry /pulsar/functions to correct IPs everything works OK.

It seems that there is an initialisation issues somehow.

Are you willing to submit a PR?

github-actions[bot] commented 1 year ago

The issue had no activity for 30 days, mark with Stale label.

ntilivea commented 1 year ago

We found the solution - there was a problem in the initialisation parameter for the ZK cluster.

when running the /opt/pulsar/bin/pulsar initialize-cluster-metadata the parameter that initialise the metadata store --metadata-store & --configuration-metadata-store need to be set to actual IPs of the cluster ( not localhost as we were setting them ).